基于容忍度的网络拓扑自适应差分进化算法  

Tolerance-based adaptive differential evolution algorithm with network topology

作　　者：李伟[1] 孙亚峰 黄颖[2] 颜雪松 LI Wei;SUN Yafeng;HUANG Ying;YAN Xuesong(School of Information Engineering,Jiangxi University of Science and Technology,Ganzhou 341000,China;School of Mathematical and Computer Sciences,Gannan Normal University,Ganzhou 341000,China;School of Computer Science,China University of Geosciences,Wuhan 430740,China)

机构地区：[1]江西理工大学信息工程学院,江西赣州341000 [2]赣南师范大学数学与计算机科学学院,江西赣州341000 [3]中国地质大学(武汉)计算机学院,湖北武汉430740

出　　处：《计算机集成制造系统》2022年第11期3479-3493,共15页Computer Integrated Manufacturing Systems

基　　金：国家自然科学基金资助项目(61903089,62066019);江西省自然科学基金资助项目(20202BABL202020,20202BAB202014);国家重点研发计划资助项目(2020YFB1713700)。

摘　　要：为了进一步提高差分进化算法的局部搜索能力,提出一种基于容忍度的网络拓扑自适应差分进化算法。通过构建最近邻耦合网络拓扑和小世界网络拓扑,采用基于容忍度的拓扑选择机制为所有个体选择网络拓扑及邻域,从邻域中选择个体参与变异操作以提升算法局部搜索性能。此外,为了使初始种群在搜索空间内的分布更加均匀,设计了一种边界反向映射初始化策略,用以替代原始的初始化策略。为验证所提策略的有效性,将所提算法与几种先进的改进差分进化算法在25个测试函数上进行了比较,实验结果表明所提算法的求解精度和收敛速度优于其他算法,具有极强的竞争力。To improve the local search ability of the differential evolution algorithm,a tolerance-based adaptive differential evolution algorithm with network topology was proposed.The nearest-neighbor coupled network topology and the small-world network topology of the population were constructed before the mutation operation.The tolerance-based topology selection mechanism selected the appropriate network topology and neighborhood for each individual.Individuals involved in the mutation were chosen from the neighborhood to enhance the export capacity.To make the population uniform well,a boundary reverse mapping initialization strategy was also designed to replace the original initialization strategy.The proposed algorithm was compared with several advanced differential evolution algorithms by using 25 test functions.The experimental result verified that the proposed algorithms outperformed the competitors in accuracy and convergence rate with extremely competitive.

关 键 词：差分进化算法 网络拓扑 最近邻耦合网络 小世界网络 初始化策略 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程] TP191[自动化与计算机技术—控制科学与工程]